Closed balhoff closed 2 years ago
For example, right now hydron has reference count 5
. But in the updated counting it would be 136
, giving a much lower IC.
We've been using it experimentally in CAP, but I don't think it's turned out to be that useful for what we were trying to achieve. @ubyndr - do you have any details to add?
We tried to utilize this to propose a grouping for their anatomical structures, but they weren't so eager to use this analysis.
Right now the precomputed information content score is based on incoming
subClassOf
andpart_of
relations: https://github.com/INCATools/ubergraph/blob/b6c07bd169aaa22fdaf376fbe2ca40cbb9ba44fb/ic.dl#L18-L19We are thinking of changing this to reflect any relation in the redundant graph. This way any term that is frequently referenced in axioms would get a lower IC score.
@dosumis is your group using this number for anything? Trying to see if anyone outside of Data Translator (@cbizon) would notice this.